Claims:We Claim:
1. A method (200) of training an artificial intelligence (AI) model to detect a degree of clogging of a filter (103) in a Selective Catalytic Reduction (SCR) system, said SCR system comprising a pump (102) unit, said filter (103) , a set of sensors and at least an Electronic Control Unit (ECU (104) ), the filter (103) delivering an aqueous solution to the pump (102) unit, the set of sensors configured to measure pump (102) operating parameters, the set of sensors in communication with the ECU (104) , the training method comprising:
measuring (201) a real-time value of a set of pump (102) operating parameters by means of the first set of sensors;
retrieving (202) values for a second set of parameters for the measured real-time values from the ECU (104) ;
feeding (203) the measured values and the retrieved values as input to the AI model;
labelling (204) the output in dependance of data received from an experimental sensor to determine the filter (103) clogging.

2. A method (200) of training an artificial intelligence (AI) model to detect a degree of clogging of a filter (103) as claimed in claim 1, wherein a set of pump (102) operating parameters comprise parameters dependent on current or voltage characteristics of the pump (102) with respect to time.

3. A method (200) of training an artificial intelligence (AI) model to detect a degree of clogging of a filter (103) as claimed in claim 1, wherein the second set of parameters comprise either one or more system operating parameters such battery voltage and at least a quantity of aqueous solution to be dosed.

4. A method (300) to predict a remaining useful life of a filter (103) in a Selective Catalytic Reduction (SCR) system, the SCR system comprising a pump (102) unit, said filter (103) , a set of sensors and at least an Electronic Control Unit (ECU (104) ), the filter (103) delivering an aqueous solution to the pump (102) unit, the set of sensors configured to measure pump (102) operating parameters, the set of sensors in communication with the ECU (104) , the method comprising:
measuring (301) a real-time value of a set of pump (102) operating parameters by means of a first set of sensors;
retrieving (302) values for a second set of parameters for the measured real-time values by the ECU (104) ;
feeding (303) the measured values and the retrieved values to a pre-trained AI model;
executing (304) the pre-trained AI model to determine a degree of filter (103) clogging;
comparing (305) the determined degree of filter (103) clogging with a pre-retrieved historical data to predict a remaining useful life of a filter (103) .

5. The method (300) to predict a remaining useful life of a filter (103) as claimed in claim 4, wherein the set of pump (102) operating parameters comprise parameters dependent on current or voltage characteristics of the pump (102) with respect to time.

6. The method (300) to predict a remaining useful life of a filter (103) as claimed in claim 4, wherein the second set of parameters comprise either one or more system operating parameters such battery voltage and at least a quantity of aqueous solution to be dosed.

7. An electronic control unit (ECU (104) ) adapted to predict a remaining useful life of a filter (103) in a Selective Catalytic Reduction (SCR) system, said SCR system comprising a pump (102) unit, said filter (103) , a set of sensors and at least an Electronic Control Unit (ECU (104) ), the filter (103) delivering an aqueous solution to the pump (102) unit, the set of sensors configured to measure a set of pump (102) operating parameters, the set of sensors in communication with the ECU (104) , the ECU (104) configured to:
Actuate the pump (102) unit;
measure a real-time value of a set of pump (102) operating parameters by means of a first set of sensors;
retrieve values for a second set of parameters for the measured real-time values;
feed the measured values and the retrieved values to a pre-trained AI model;
receive an output from the pre-trained AI model to determine a degree of filter (103) clogging;
compare the determined degree of filter (103) clogging with a pre-recorded historical data to predict a remaining useful life of a filter (103) .

8. The electronic control unit (ECU (104) ) adapted to predict a remaining useful life of a filter (103) as claimed in claim 7, wherein the set of pump (102) operating parameters comprise parameters dependent on current or voltage characteristics of the pump (102) with respect to time.

9. The electronic control unit (ECU (104) ) adapted to predict a remaining useful life of a filter (103) as claimed in claim 7, wherein the second set of parameters comprise either one or more system operating parameters such battery voltage and at least a quantity of aqueous solution to be dosed.
, Description:Complete Specification:
The following specification describes and ascertains the nature of this invention and the manner in which it is to be performed

Field of the invention
[0001] The present disclosure relates to a method of training an artificial intelligence (AI) model and detecting a degree of clogging of a filter in a Selective Catalytic Reduction (SCR) system, thereby predicting the remaining useful lifetime and enable effective trouble shooting and enhance/adapt the system functioning by understanding the degree of deterioration.

Background of the invention
[0002] Selective Catalytic Reduction (SCR) is an emissions control technology system used in diesel engines. In this technology a liquid reductant agent (aqueous solution) is injected into the exhaust stream of a diesel engine. The liquid reductant agent is usually automotive-grade urea, otherwise known as Diesel Exhaust Fluid (DEF). The DEF sets off a chemical reaction that converts Nitrogen Oxides (NOx) into nitrogen, water and tiny amounts of carbon dioxide (CO2), which is then expelled through the vehicle tailpipe.

[0003] DEF is stored in a tank and through a pump unit delivered to a dosing module which doses it into the exhaust stream. The DEF filter is installed in SCR module to ensure the DEF does not have contaminants that deteriorate the pump, or the injector and the DEF dosed into the exhaust gas stream is free from contaminants. After prolonged use the DEF filter gets clogged and needs replacement. There is a need for vehicle user to know the degree to which this DEF filter is clogged for effective predictive diagnosis. With the advancements in data science future outcomes such as failure of a component can be predicted using historical data combined with statistical modeling, data mining techniques and machine learning. Hence there is a need to apply predictive analytics in the SCR system to determine remaining useful life of the filter.

[0004] Patent application IN202041006256AA titled “Method to detect clogging in a fuel filter of a vehicle” discloses a method of detecting a degree of fuel filter clogging using a self-learning algorithm. In step 201, an ECU (105) receives a measured values of a set of parameters. In step 202, the ECU (105) stores a matrix comprising a correlation factor between each of the set of parameters and value of differential pressure. In step 203, the ECU (105) calculates a predicted value of differential pressure based on the measured values of the set of parameters and the matrix. In step 204, the ECU (105) linearizes the predicted value of differential pressure with a dynamic value of actual differential pressure. The dynamic value of actual differential pressure is derived from a self-learning algorithm using the actual value of differential pressure measured at various instances. In step 205, the ECU (105) indicates the value of degree of filter clogging to the vehicle user.

Brief description of the accompanying drawings
[0005] An embodiment of the invention is described with reference to the following accompanying drawings:
[0006] Figure 1 depicts a portion (100) of a Selective Catalytic Reduction (SCR) system in a vehicle;
[0007] Figure 2 illustrates method steps (200) for training an artificial intelligence (AI) model to detect a degree of clogging of a filter (103) in a Selective Catalytic Reduction (SCR) system; and at least
[0008] Figure 3 illustrates method steps (300) for predicting a remaining useful life of a filter (103) in a Selective Catalytic Reduction (SCR) system.

Detailed description of the drawings

[0009] Figure 1 depicts a portion (100) of an SCR system. The SCR system comprises a diesel exhaust fluid (DEF) tank (101) , DEF filter (103) and at least a pump (102) unit amongst the other downstream components such a dosing module and other components known to a person skilled in the art. The DEF tank (101) stores an aqueous solution known as diesel exhaust fluid or adblue. The DEF filter (103) is located downstream from the DEF tank (101) and filter (103) s the contaminants in the ad blue. The filter (103) delivers the aqueous solution to the pump (102) unit. The pump (102) unit located downstream from the DEF filter (103) supplies the diesel exhaust fluid to the dosing module. The dosing module regulates the amount of DEF directed into exhaust gas pipe. One or more sensors are configured to measure pump (102) operating parameters. These sensors are in communication with an electronic control unit (ECU (104) ).The set of sensors configured to measure pump (102) operating parameters.

[0010] The ECU (104) is a combination of: one or more microchips or integrated circuits interconnected using a parent board, hardwired logic, software stored by a memory device and executed by a microprocessor, firmware, an application specific integrated circuit (ASIC), and/or a field programmable gate array (FPGA). The ECU (104) is in communication with a artificial intelligence (AI) model. In an exemplary embodiment the AI module may be a part of the ECU (104) . The AI module can either be a software embedded in a single chip or a combination of software and hardware where each module and its functionality is executed by separate independent chips connected to each other to function as the system. For example, a neural network chips that are specialized silicon chips, which incorporate AI technology and are used for machine learning can be embedded within the ECU (104) .

[0011] It should be understood at the outset that, although exemplary embodiments are illustrated in the figures and described below, the present disclosure should in no way be limited to the exemplary implementations and techniques illustrated in the drawings and described below.

[0012] Figure 2 illustrates method steps for training an artificial intelligence (AI) model to detect a degree of clogging of a filter (103) in a Selective Catalytic Reduction (SCR) system. The SCR system is the same as explained in accordance with figure 1. It is reiterated that said SCR system comprises a pump (102) unit, said filter (103) , a set of sensors and at least an Electronic Control Unit (ECU (104) ). The filter (103) delivers an aqueous solution to the pump (102) unit, the set of sensors are configured to measure pump (102) operating parameters, the set of sensors in communication with the ECU (104) . During the training phase, a differential pressure sensor is employed across the filter (103) to fetch real time pressure drop across the filter (103) . The AI model to be trained in an exemplary embodiment, resides within the ECU (104) . In other embodiments the AI model can be separate from the ECU (104) but as a part of the said SCR system i.e. in communication with the ECU (104) . The method steps (200) are explained in accordance with an exemplary embodiment of the present disclosure.

[0013] Method step 201 comprises measuring a real-time value of a set of pump (102) operating parameters by means of the first set of sensors. The set of pump (102) operating parameters comprise parameters dependent on current or voltage characteristics of the pump (102) with respect to time. For example these can be value of current (iBMP) at the begin of motion point (BMP), value of current (iMSP) at the Mechanical stop point (MSP), time taken (tiBMP) to reach the BMP, time taken (tiMSP) to reach the MSP. Likewise, BSP and ESP points which correspond to the suction stroke in similar terms.

[0014] BMP is the point in time when the spool displacement starts in the actuation stroke. MSP is the point in time when the spool displacement stops in the actuation stroke. The actuation stroke is the result of solenoid energizing in the pump armature. BSP is the point in time when the spool displacement starts in the suction stroke. ESP is the point in time when the spool displacement stops in the suction stroke. The suction stroke is the result of potential energy of the coiled spring in the pump armature.

[0015] Method step 202 comprises retrieving values for a second set of parameters for the measured real-time values from the ECU (104) . The second set of parameters comprise either one or more system operating parameters such battery voltage and at least a quantity of aqueous solution to be dosed. If needed for accuracy other parameters such as ambient temperature and/or pressure can be included to train the model.

[0016] Method step 203 comprises feeding the measured values and the retrieved values as input to the AI model. In an embodiment of the present disclosure the measured and retrieved values fed to the AI model with a moving average using a window length of 30 samples at different levels of inlet side pressure drop. The type of AI model could be chosen from the group linear regression, K Nearest Neighbours, Random Forest Classifier, Decision Tree Classifier, naïve bayes classifier, support vector machine, neural networks, and the like.

[0017] Method step 204 comprises labelling the output in dependance of data received from an experimental sensor to determine the filter (103) clogging. During the training phase additionally a differential pressure sensor is employed to fetch real time pressure drop across the filter (103) , thereby indicating clogging. While training the AI model, for example a 70%-30% train-test split of dataset is used. Meaning 70% of the time the output is labelled with the actual value from the differential pressure sensor and rest of the 30% it is left back-propagate and learn the value itself based the previous inputs and labelled outputs. The accuracies obtained by various models on the tests and the best suited model chosen or the model is sent for retraining with more diverse input until an optimal accuracy is exhibited by the AI model.

[0018] Figure 3 illustrates method steps for predicting a remaining useful life of a filter (103) in a Selective Catalytic Reduction (SCR) system. The SCR system is the same as explained in accordance with figure 1. It is reiterated that said SCR system comprises a pump (102) unit, said filter (103) , a set of sensors and at least an Electronic Control Unit (ECU (104) ). The filter (103) delivers an aqueous solution to the pump (102) unit, the set of sensors are configured to measure pump (102) operating parameters, the set of sensors in communication with the ECU (104) . Further in an exemplary implementation of the said methods steps (300), a pre-trained AI model resides within the ECU (104) . The AI model has been trained in accordance with method steps (200) of figure 2.

[0019] Method step 301 comprises measuring a real-time value of a set of pump (102) operating parameters by means of a first set of sensors. The first set of sensors measure the set of pump (102) operating parameters dependent on current or voltage characteristics of the pump (102) with respect to time. These values are measured continuously by the sensors and the data in continuously sent to the ECU (104) . For example these can be value of current (iBMP) at the begin of motion point (BMP), value of current (iMSP) at the Mechanical stop point (MSP), time taken (tiBMP) to reach the BMP, time taken (tiMSP) to reach the MSP. Likewise, BSP and ESP points which correspond to the suction stroke in similar terms.

[0020] BMP is the point in time when the spool displacement starts in the actuation stroke of the pump. MSP is the point in time when the spool displacement stops in the actuation stroke. The actuation stroke is the result of solenoid energizing in the pump armature. BSP is the point in time when the spool displacement starts in the suction stroke. ESP is the point in time when the spool displacement stops in the suction stroke. The suction stroke is the result of potential energy of the coiled spring in the pump armature.

[0021] Method step 302 comprises retrieving values for a second set of parameters for the measured real-time values by the ECU (104) . The second set of parameters comprise either one or more system operating parameters such battery voltage and at least a quantity of aqueous solution to be dosed. It can be expanded to include other parameters such as ambient temperature and/or pressure etc.

[0022] Method step 303 feeding the measured values and the retrieved values to a pre-trained AI model. The AI model has been trained in accordance with method steps (200). The type of AI model could be chosen from amongst linear regression, K Nearest Neighbours, Random Forest Classifier, Decision Tree Classifier, naïve bayes classifier, support vector machine, neural networks, and the like in accordance with the system requirements and the accuracy exhibited during training.

[0023] Method step 304 comprises executing the pre-trained machine learning model to determine a degree of filter (103) clogging. The AI model runs the trained algorithm to detect a degree of pressure drop across the filter (103). The detected pressure drop is correlated with the degree of filter (103) clogging based on historical data analysis and the filter (103) characteristics.

[0024] Method step 305 comparing the determined degree of filter (103) clogging with a pre-retrieved historical data to predict a remaining useful life (RUL) of a filter (103) . The RUL of the filter (103) can be predicted using a history of filter (103) performance data obtained from the AI model. The rate at which the filter (103) is deteriorating is constantly tracked and stored as historical data. This can be further extended by extrapolation to determine the RUL of the filter (103) in terms of time, the number of kilometers, volume of DEF consumption, number of refill events, etc. The same can be indicated to the user by an audio or visuals means on the dashboard.

[0025] A person skilled in the art will appreciate that while these method steps describes only a series of steps to accomplish the objectives, these methodologies may be implemented by a single ECU (104) or a combination of a few. The ECU (104) should be adapted to actuate the pump (102) unit; measure a real-time value of a set of pump (102) operating parameters by means of a first set of sensors; retrieve values for a second set of parameters for the measured real-time values; feed the measured values and the retrieved values to a pre-trained AI model; receive an output from the pre-trained AI model to determine a degree of filter (103) clogging; compare the determined degree of filter (103) clogging with a pre-recorded historical data to predict a remaining useful life of a filter (103) .

[0026] In the experimentation and testing phase various AI models such as K Nearest Neighbors, Decision Tree Classifier, Random Forest Classifier were tested in the experimental set-up. The test setup used a throttle valve to replace the pressure drop created by the filter (103) . The pressure drop developed was controlled using the extent of throttling produced by the valve. Pressure sensors were employed on either side of the throttle valve to monitor pressure drop. An external current clamp was used to record continuous pump (102) current values. Hydraulic head loss due to the various pipes and fittings used have been calculated and deviations from the real-world working conditions of the system have been compensated in the AI model. The testing data has been recorded at varying levels of pump (102) inlet side pressure drop, pump (102) voltage as well as dosing quantities. The AI models such K Nearest Neighbors, Decision Tree Classifier, Random Forest Classifier exhibited accuracies ranging from 90 -96%.

[0027] This idea to develop a method of training an artificial intelligence (AI) model and determining remaining useful life of a filter (103) using the trained model thereof provides the user with a real time deterioration data of this filter (103) . Additional Inferences can be drawn using the invention such as comparative assessment of DEF quality based on the rate of filter (103) performance deterioration. Furthermore, pressure build-up error or under pressure error arising due to either a leaky pressure line, faulty pump (102) or filter (103) deterioration can be ascertained using the AI model when it arises due to filter (103) deterioration because of clogging. It also leads to enhancement of system functioning like volumetric corrections, additional strokes, change in frequency etc. can be implemented.

[0028] It must be understood that the embodiments explained in the above detailed description are only illustrative and do not limit the scope of this invention. Any modification to a method of training an artificial intelligence (AI) model and determining remaining useful life of a filter (103) are envisaged and form a part of this invention. The scope of this invention is limited only by the claims.